Your words, less work.
Talk naturally. VOCA Polish runs Qwen 4B directly on Apple Silicon to tidy your words with your chosen style. Download once; enhance offline. No API key or companion app.
Think it. Say it. Let Voca do the typing.
Thoughtful voice dictation, right where you write.
Hey Alex, let’s take the meeting outside.
I think our best ideas could use a little fresh air.
Right where your words belong.
No new workflow. No blank-page staring contest.
Just your voice, and a little help from Voca.
“Hey Alex, um, let’s take the meeting outside. I think, uh, our best ideas could use a little fresh air.”
Talk naturally. VOCA Polish runs Qwen 4B directly on Apple Silicon to tidy your words with your chosen style. Download once; enhance offline. No API key or companion app.
An email, a big idea, a message you’ve been putting off. It all starts with one key, right where you’re working.
A discreet little companion that stays out of the way. No extra window to babysit. Just room to think.
VOCA’s character is simple: give you room to think, keep you in control, and never make you repeat a good thought because a server is having a moment.
Start with 12 writing templates, or supply your own writing samples. VOCA finds sentence and punctuation habits locally, then lets you review the style. No background typing surveillance. Your personality is not a telemetry event.
Optional automatic finish waits 3, 7, or 12 seconds, then shows a three-second countdown. Speak again to reset it; tap the countdown to keep listening. Quiet is a clue, not mind reading.
Check insertion in your chosen app. Undo a verified insertion without removing later appends. If the field changes in a way VOCA can’t safely reverse, it refuses the edit and offers your words to copy.
A compact keyboard-accessible chooser for your model, microphone, and writing style. Your destination icon stays beside the waveform, at the notch, screen edge, or near the typing cursor.
Slow dictation cleanup gets a warning after eight seconds. Keep waiting, use the original words now, or let the thirty-second limit do it for you. Late AI answers don’t get a second insertion.
Benchmark this Mac and time an installed speech model. Keep the full model library and your own cleanup providers. See time saved and recent activity without a trophy cabinet demanding attention.
*Parakeet TDT v3, 1.2-second bundled sample, median of three warm runs on a 16 GB, 10-core Apple Silicon Mac. Measured September 11, 2026. This is not an accuracy score, a competitor benchmark, or a promise for every recording. Cursor placement and precise undo depend on the destination app’s Accessibility support.
Because apparently the world needed
one more dictation app.
Just not another monthly habit.
Free to use. Open source. More room for your words.
Open-source dictation. Every local feature included.
Apple Silicon · macOS 15+ · About 20 MB download. Speech and AI models download separately; local Qwen adds about 2.28 GB. Liquid Glass requires macOS 26.
Before you download: this beta is locally signed, not Apple-notarized. macOS may require approval in System Settings → Privacy & Security. Microphone and Accessibility setup are required. Compatibility varies by app and model.
Download free beta ↗Installation guide, checksums & release notes ↗Matching source (GPLv3) ↗Free to use and free to build. If VOCA helps, optional donations can support development. Every feature stays available without donating.
Less mystery.
More getting on with it.
Voca turns your voice into writing on your Mac. Place your cursor in a text field, use your dictation shortcut, and speak. Your words appear where you were about to type.
Voca is designed for the places you already write: messages, emails, notes, documents, and text fields in your favorite Mac apps. The destination checker can verify a draft in your chosen app. Some custom or protected fields do not support insertion or precise undo; VOCA keeps a recovery copy when it can’t verify a safe edit.
Yes. The prebuilt beta and source are free, with all local features included. Donations are optional and never unlock features. Cloud API usage is billed separately by whichever provider you choose.
Yes. Play the interactive walkthrough above to see the journey from natural speech to finished writing. It’s a guided illustration, so it won’t use your microphone or record your voice.
No app can know that from silence alone. VOCA’s optional pause detector waits generously, adds extra time for some unfinished English phrases, and lets speech reset the countdown. Keep it off for long pauses or noisy rooms. Automatic finish never submits a message.
With an on-device speech model and local cleanup, processing stays on your Mac after model downloads. Choosing a cloud provider sends the required content to that provider under its terms. Personal-style examples are analyzed locally; only the saved style instructions accompany future cleanup requests. History and saved audio have separate controls.
VOCA builds on FluidVoice under GPLv3, with a redesigned interface and additional workflows. The source is public on GitHub, and corresponding source accompanies distributed builds. You can inspect, build, modify and redistribute the covered code under GPLv3.
Yes, as a free beta for Apple Silicon and macOS 15+. It is locally signed but not Apple-notarized, so macOS may require manual approval in Privacy & Security. Read the release’s installation guide first. Testing across more Macs, languages and destination apps is ongoing.
Voca is being built for Mac. This page is focused on the Mac experience; other platforms haven’t been announced.
Your keyboard can take it from here. Or take a break.